Beruflich Dokumente
Kultur Dokumente
Kenneth F. Kroner
Managing Director
BlackRock
Laurence B. Siegel
Research Director, Research Foundation of CFA Institute
Senior Advisor, Ounavarra Capital LLC
The equity risk premium (ERP) is almost certainly the most important variable
in finance. It tells you how much you need to save, how much you can spend,
and how to allocate your assets between equities and bonds. Yet, recognized
experts cannot agree on the ERP’s value within an order of magnitude or even
agree whether it is negative or positive. At a 2001 symposium, the predecessor
of the one documented in this book, Robert Arnott and Ronald Ryan set forth
an ERP estimate of –0.9 percent and Roger Ibbotson and Peng Chen proposed
+6 percent.1 The estimates in this book are much more tightly clustered, but
considerable disagreement remains about how to estimate the premium as well
as its size.
Grinold and Kroner (2002) proposed a model of the ERP that linked equity
returns to gross domestic product (GDP) growth.2 The key insight, which
draws on earlier work by a number of authors, was that aggregate corporate
profits cannot grow indefinitely much faster—or much slower—than GDP.
(And as Herbert Stein was fond of reminding us, any economic trend that
cannot continue forever will not.) If profits grow faster than GDP, they
eventually take over the economy, leaving nothing for labor, government,
natural resource owners, or other claimants. If profits grow more slowly than
1 See Arnott and Ryan (2001); Ibbotson and Chen (2003). The Ibbotson and Chen estimate of
6 percent is an arithmetic mean expectation; their geometric mean expectation was 4 percent.
2 A second printing of this article, from March 2004, is available online at www.cfapubs.org/
userimages/ContentEditor/1141674677679/equity_risk_premium.pdf.
GDP, they eventually disappear and businesses will have no profit motive to
continue operating. Thus, in the very long run, the ratio of profits to GDP is
roughly constant.
The title of this paper, a shortened and updated version of Grinold and
Kroner (2002), refers to the “supply model” of Diermeier, Ibbotson, and Siegel
(1984), who differentiated between the demand for capital market returns (what
investors need to compensate them for risk) and the supply of returns (what the
macroeconomy makes available). The original supply model likewise made use
of a link between profits and GDP. Grinold and Kroner (2002) was titled “The
Equity Risk Premium: Analyzing the Long-Run Prospects for the Stock
Market,” but the similarity with the title of this book forced us to rename the
current paper. Although our method is designed to produce an ERP estimate
that reflects both supply and demand, the link to macroeconomic performance
gives it a supply-side flavor.3
When we revisited the estimates from Grinold and Kroner (2002), we
found that not all the components could be updated with equal accuracy, so the
ERP estimate provided here is subject to some important caveats regarding data
adequacy. The method that we recommend, however, remains largely
unchanged from Grinold and Kroner (2002).
The return to equities over a single period can always be broken down as
RS Income return Nominal earnings growth Repricing. (2)
The income return is the percentage of market value that is distributed to
shareholders as cash. If dividends are the only source of income, then the income
return is equivalent to the dividend yield. Today, share repurchase programs
(buybacks) are another common means of distributing cash to shareholders.
Cash takeovers (by one company of another) should also be counted in the
income return of an index that includes the stock of the acquired company.
The next two terms in Equation 2 represent the capital gain. Capital gains
come from a combination of earnings growth and P/E expansion or contraction,
which we call “repricing.”
For expository purposes, we decompose the components further and use
more precise notation. The return over a single period is
D
R S i g PE .
P
(3)
Income Earnings growth Repricing
The first term, D/P, is simply the dividend yield. The second term, –S,
is the percentage change in the number of shares outstanding. The percentage
change in the number of shares outstanding equals the “repurchase yield”
(which theoretically also includes cash takeovers) minus new shares issued
(dilution); it has a negative sign because a decrease in the number of shares
outstanding adds to return and an increase subtracts from return.4 Together,
the terms D/P and –S measure the fraction of market capitalization that the
companies in an index, in aggregate, return to shareholders in cash. Therefore,
we refer to the sum of these two terms as the “income return.”
The remaining terms, i + g + PE, make up the capital gain. The term i
represents the inflation rate. The term g is the real earnings (not earnings per
share) growth rate over the period of measurement. The final term, PE, is
the percentage change in the P/E multiple over the period. We refer to this
last piece as the “repricing” part of the return.
4 Share buybacks may be viewed as either a component of income return or a component of capital
gain. An owner of a single share who holds on to the share through the share buyback program
experiences the buyback as a component of capital gain because the same earnings are divided
among fewer shares, which causes EPS to rise although earnings (not per share) have not
changed. If the stock’s P/E and all other factors are held equal, then the stock price rises. An
index fund investor, however, experiences the share buyback as cash income because the index
fund manager—who tenders some of the shares to the issuer to keep the stock’s (now decreased)
weight in the fund proportionate to its weight in the index—receives cash, which is then
distributed to, or held by, fund shareholders like any other cash (tax considerations aside). We
choose to view share buybacks as a component of income return.
Historical Returns
Let us briefly consider what risk premium markets have provided historically.
Over the last 85 years (1926–2010), the U.S. stock market and the intermediate-
term U.S. Treasury bond market have delivered compound annual nominal
returns of 9.9 percent and 5.4 percent, respectively.5 Thus, the realized pre-
mium that stocks delivered over bonds was 4.5 percent.6 The historical return
decomposition in Table 1 can be used to better understand this 9.9 percent
annual equity return.
The income return (through dividends only, not share buybacks) on the
S&P 500 was 4.1 percent annualized over this 85-year period. In this decom-
position, we adjusted earnings growth for increases in the number of shares to
arrive at earnings per share (EPS) growth. EPS grew at a rate of about 4.9 percent
per year (1.9 percent real growth and 3.0 percent inflation) over the period.
5 See the data for large-company stocks (i.e., the S&P 90 from January 1926 through February
1957 and the S&P 500 thereafter) in Table 2.1 in Ibbotson SBBI (2011, p. 32). Returns are
before fees, transaction costs, taxes, and other costs.
6 This amount is the arithmetic difference of geometric means. The geometric difference of
geometric means, or the compound annual rate at which stocks outperformed bonds, is given by
(1 + 0.099)/(1 + 0.054) – 1 = 4.27 percent.
The remainder of the total return on equities was due to repricing. The
P/E of the market, measured as the end-of-year price divided by trailing
12-month earnings, grew from 11.3 at year-end 1925 to 18.5 at year-end
2010.7 This repricing works out to an additional return, or P/E expansion, of
0.58 percent per year. A common view is that this P/E expansion was
understandable and reasonable in light of the technological and financial
innovations over this long period. For example, accounting standards became
more transparent (recent “fraud stocks” notwithstanding). Such innovations
as the index fund made it easier for investors to diversify security-specific risk
and to save on costs. Mutual fund complexes provided easier access to institu-
tional-quality active management. Finally, many market observers perceive the
business cycle to have been under better control in recent decades than it was
in the 1920s and 1930s, which made expected earnings smoother; the recent
near depression and quick recovery, at least in corporate profits and the stock
market, support this view somewhat. All these factors have made equity
investing less risky and contributed to the repricing over this 85-year period.
But the presence of these factors in the past does not mean that we should
build continued upward repricing into our forecasts. We consider this issue later
in this paper.
Chart 1 of Grinold and Kroner (2002) further dissects the return decom-
position into annual return contributions. Their graph demonstrates that the
noisiest component of returns is clearly P/E repricing, followed by real earnings
growth. Inflation and income returns are relatively stable through time. This
observation implies that our real earnings growth and repricing forecasts are
likely to be the least accurate and our inflation and income return forecasts are
likely to be more accurate.
Mehra and Prescott (1985), and many others, argued that the equity
premium of 4.5 percent was a multiple of the amount that should have been
necessary to entice investors to hold on to the risky cash flows offered by equities
instead of the certain cash flows offered by bonds. This contention spawned a
huge literature on the “equity risk premium puzzle.”8 We have always been
perplexed by a debate that suggests that investors were wrong while a specific
macroeconomic theory is right, but Rajnish Mehra sheds additional light on
this question elsewhere in this book.
7 Because earnings were growing very quickly at the end of 2010, the more familiar P/E calculated
as the current price divided by 12-month forward (forecast) earnings was lower than the P/E
shown here.
8 For surveys of this literature, see Kocherlakota (1996); Mehra (2003).
It is possible to make the case for a much higher repurchase yield forecast
by giving greater weight to more recent information (which is basically what
we did with the dividend yield). According to Standard & Poor’s (2008), “Over
the past fourteen quarters, since the buyback boom began during the fourth
quarter of 2004, S&P 500 issues have spent approximately $1.55 trillion on
stock buybacks compared to . . . $783 billion on dividends.” Although buybacks
collapsed in 2009, they rebounded in 2010 and 2011. If the two-to-one ratio
of buybacks to dividend payments observed by Standard & Poor’s over 2004–
2008 persists in the future, the repurchase yield will be as high as 3.5–3.6
percent. Aiming for a “fair and balanced” estimate, we use the lower number,
2.2 percent, which we obtained by weighting all 12 years of historical share
repurchase data equally.11
We have not included cash buyouts in our estimate of the repurchase yield.
From the perspective of an investor who holds an index containing companies
A, B, C, and so forth, a cash buyout or takeover—a payment by company A to
11 The use of this lower number is neutral, not conservative in the sense of numerically
minimizing the ERP estimate. The reason is that there are offsetting biases. Our buyback
estimate of 2.2 percent is too high because we do not subtract the historical contribution of
buybacks to the dilution estimate (discussed later). And it is too low because very recent buyback
rates have been much higher than 2.2 percent, not to mention the fact that we fully ignore the
cash takeover yield.
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0
1900 10 20 30 40 50 60 70 80 90 2000 10
We expect real earnings growth to equal real GDP growth for the macro-
consistency reason stated earlier: Any other result would, in the very long run,
lead to an absurdity—corporate profits either taking over national income
entirely or disappearing. Figure 2 shows the (trendless) fluctuations in the
corporate profit share of GDP since 1947.
These observations leave us with the puzzle of forecasting real GDP
growth. Grinold and Kroner (2002) engaged in a fairly typical macroeconomic
analysis that involved productivity growth, labor force growth, and the expected
difference between S&P 500 earnings and overall corporate profits. They did
not use historical averages or trends directly as forecasts; rather, they argued
that the data plus other factors justified the conclusion that real GDP would
most likely grow at 3 percent over the relevant forecast period and that real S&P
500 earnings would grow at 3.5 percent.
Real economic growth, by definition, equals real productivity growth plus
labor force growth. Although we can update the historical productivity and
labor force growth numbers, doing so would not produce an especially useful
forecast any more than it did for Grinold and Kroner (2002), who distanced
themselves somewhat from the productivity and labor force growth approach.
The reason is that extrapolating recent trends in these components of eco-
nomic growth can produce unrealistically high or low expectations, and using
Profits/GDP (%)
15
13
11
5
47 50 53 56 59 62 65 68 71 74 77 80 83 86 89 92 95 98 01 04 07 10
Figure 3. U.S. Real Productivity and Labor Force Growth Rates, 1971–2009
–1
71 73 75 77 79 81 83 85 87 89 91 93 95 97 99 01 03 05 07 09
Source: Organisation for Economic Co-Operation and Development, OECD StatExtracts (http://
stats.oecd.org/Index.aspx, as of 14 November 2011: total labour force, U.S., and labour productivity
annual growth rate, U.S.).
10,000
1,000
100
1789 1819 1849 1879 1909 1939 1969 1999
We add to the 1.8 percent real per capita GDP growth estimate the
Economist Intelligence Unit 10-year U.S. population growth estimate of 0.85
percent,15 which gives a total real GDP growth forecast of 2.65 percent. This
number is slightly below current consensus estimates.
This simplified method presents some difficulty because if the rate of
dilution is 2 percent at all population growth rates, then population growth has
a one-for-one effect on the estimate of the expected return on equities and,
therefore, on the ERP. This suggests an easy beat-the-market strategy: Invest
only in countries with the fastest population growth. This strategy has not
worked well in the past, and even if it did over some sample period, easy beat-
the-market strategies are usually illusory. Thus, the dilution estimate should
probably be higher for countries with high population growth rates or for a
country during periods of above-normal population growth. Although the logic
of using a link to real GDP growth to forecast the stock market has great
intuitive appeal, putting it into practice with any precision will take more work
and more thought regarding dilution.16
Expected Inflation. Because we are deriving the ERP relative to
Treasury bonds, we do not need our own inflation forecast as much as we need
an estimate of the inflation rate that is priced into the 10-year Treasury bond
market. Historical inflation rates have no bearing on this number, so we do
not present them. Fortunately, the yield spread between 10-year nominal
Treasury bonds and 10-year TIPS is a direct, although volatile, measure of the
inflation rate that is expected by bondholders. (The spread also includes an
inflation risk premium, present in nominal bond yields but not in TIPS yields,
for which we need to adjust.)
15 This number was obtained at http://7marketspot.com/archives/2276 on 2 May 2011 under
the heading “USA economy: Ten-year growth outlook” in the column “2011–20.” If we instead
used real productivity growth plus labor force growth to estimate real GDP growth, we would
get a slightly higher number for real productivity growth and a slightly lower number for labor
force growth, which would provide a very similar overall real GDP forecast.
16 Our simplified method has some other characteristics worth noting. It does not specifically
account for the wedge between population growth and labor force growth if the proportion of
retirees (or children) in the population is expected to change. A growing unproductive retiree
population should be considered bearish. Many would-be retirees, however, are not financially
prepared for retirement and, willingly or not, will work longer than they originally anticipated,
which contributes to GDP. In addition, in an advanced technological society, an aging
population distribution within the workforce is not all bad! We are accustomed to thinking of
young workers as productive and older workers as unproductive, but this is the case only in a
fairly primitive economy where the primary job description is something like “lift this and put it
over there.” In a technological society, young workers are unproductive—often startlingly so,
earning only the minimum wage—and older workers produce most of the added value and make
the lion’s share of the money. Nevertheless, young workers’ productivity grows quickly and older
workers’ productivity grows slowly or shrinks, so the impact of an aging workforce on rates of
change in productivity may be less salutary than the impact on the level of productivity.
On 22 April 2011, the breakeven inflation rate (the yield spread described
above) was 2.60 percent.17 This rate is high by recent standards—it was as low
as 1.5 percent in September 2010—but it is typical of the longer history of the
series. Recent concerns about very high and rapidly growing levels of public
indebtedness (of the U.S. government, of local governments in the United States,
and of non-U.S. governments) have contributed to the increase in inflation
expectations. We subtract 0.2 percent for the inflation risk premium to arrive at
a 2.4 percent compound annual inflation forecast over the next 10 years.18
Expected Repricing. Grinold and Kroner (2002, p. 15, Chart 8)
conducted an analysis of the market’s P/E that led them to include a nonzero
(–0.75 percent per year) value for the repricing term, PE, in Equation 3. At
the time the analysis was conducted (November 2001), the market’s conven-
tional trailing P/E (price divided by one-year trailing earnings) was a lofty 29.7
and the “Shiller P/E” (price divided by 10-year trailing real earnings) was 30.0,
which prompted the authors to conclude that the P/E was likely to decline.19
(The Shiller P/E is designed to smooth out fluctuations caused by yearly
changes in earnings.) And decline it did.
Today, the situation is different. Figure 5 shows the conventional P/E and
the Shiller P/E of the U.S. market. Today’s conventional P/E of 18.5 is only
modestly higher than the very long-run (1900–2010) average P/E of 15.7, and
it is lower than the more recent long-run (1970–2010) average P/E of 18.9.
The Shiller P/E tells a slightly less favorable story: The current value is 22.4,
compared with an average of 16.3 over 1900–2010 and 19.2 over 1970–2010.20
Because it averages 10 years of trailing earnings, however, the current Shiller
P/E includes an earnings collapse in 2008–2009 that is almost literally unprec-
edented; even the Great Depression did not see as sharp a contraction in S&P
composite index earnings, although overall corporate profits in 1932 were
negative. (Huge losses in a few large companies, such as those that occurred in
2008–2009, go a long way toward erasing the profits of other companies when
summed across an index.) Only the depression of 1920–1921 is comparable.
Thus, we see no justification for using a nonzero value for the repricing term
in Equation 3. The market’s current level is already reflected in the (low)
dividend yield. To include a repricing term even though the dividend yield
already incorporates the market’s valuation is, theoretically, not double-counting
because the influence of the dividend yield is amortized over an infinite horizon,
17 See www.bloomberg.com/apps/quote?ticker=USGGBE10:IND.
18 This estimate of the inflation risk premium comes from Hördahl (2008, p. 31, Graph 2).
19 Shiller (2000) describes the Shiller P/E.
20 In this section, “current” values are as of December 2010.
Figure 5. Conventional and Shiller P/Es for the U.S. Equity Market,
1900–2010
P/E
50
45
40
35 Shiller
30
25
20
15
10
5 Conventional
0
1900 10 20 30 40 50 60 70 80 90 2000 10
whereas our forecast is for only the next 10 years. Thus, if we believe that the
market is mispriced in such a way that it will be fully corrected within 10 years,
a nonzero repricing term is warranted. Although Grinold and Kroner (2002)
argued that the market P/E was too high at that time and would decline at an
expected rate of 0.75 percent per year over the forecast horizon, we think the
market is currently not too high (or too low), and our repricing forecast is zero.
2
1 rG 1 rA . (4)
2
We use standard deviations drawn from 1970 to 2010 because we do not
necessarily expect bond returns to be as placid as they have been recently. Thus,
for the purpose of estimating standard deviations, we include this long period
because it includes the bond bear market of 1970–1980 and the dramatic
subsequent recovery.22 We obtain the following:
Expected arithmetic mean equity total return = 8.59 percent.
Expected arithmetic mean 10-year Treasury bond total return = 3.96 percent.
Difference (expected arithmetic mean ERP) = 4.63 percent.
A limitation of this study is that we use U.S., not global, macroeconomic
data in our estimate of the expected return on the S&P 500. The S&P 500 is
a global index, in that it contains many companies that earn most, or a
substantial share, of their profits outside the United States. Perhaps global
economic growth rates are more relevant to the expected return on the S&P
500 than U.S. growth rates. Future research should examine this possibility.
Grinold and Kroner’s (2002) forecast, evaluated over 2002–2011, was too
high. The main problem was the volatile repricing term. They seriously under-
estimated the speed with which the unusually high P/Es that then prevailed
would revert toward their historical mean. In this paper, we forecast a repricing
of zero, consistent with our view that the market is finally, after two bear
markets and two recoveries, roughly fairly priced. Because the repricing term is
noisy, we know that our current forecast is more likely to be too high or too low
than just right when evaluated over the next 10 years. We believe, however, that
we have identified the middle of the range of likely outcomes. Although black
swans, fat tails, and tsunamis are the talk of the day, such large unexpected
events tend to fade in importance as they are averaged in with less dramatic
events over extended periods and the underlying long-term trends reveal
themselves once more.23 We expect moderate growth in the stock market.
R EFERENCES
Arnott, Robert D., and Ronald J. Ryan. 2001. “The Death of the Risk Premium.” Journal of
Portfolio Management, vol. 27, no. 3 (Spring):61–74.
Bernstein, William J., and Robert D. Arnott. 2003. “Earnings Growth: The Two Percent
Dilution.” Financial Analysts Journal, vol. 59, no. 5 (September/October):47–55.
Cornell, Bradford. 2010. “Economic Growth and Equity Investing.” Financial Analysts Journal,
vol. 66, no. 1 (January/February):54–64.
Diermeier, Jeffrey J., Roger G. Ibbotson, and Laurence B. Siegel. 1984. “The Supply of Capital
Market Returns.” Financial Analysts Journal, vol. 40, no. 2 (March/April):74–80.
Fama, Eugene F., and Kenneth R. French. 2001. “Disappearing Dividends: Changing Firm
Characteristics or Lower Propensity to Pay?” Journal of Financial Economics, vol. 60, no. 1
(April):3–43.
Fenn, George W., and Nellie Liang. 2000. “Corporate Payout Policy and Managerial Stock
Incentives.” Working paper, Federal Reserve Board (March).
Grinold, Richard C., and Kenneth F. Kroner. 2002. “The Equity Risk Premium: Analyzing the
Long-Run Prospects for the Stock Market.” Investment Insights, vol. 5, no. 3 (July):7–33.
Grullon, Gustavo, and Roni Michaely. 2000. “Dividends, Share Repurchases and the Substitu-
tion Hypothesis.” Unpublished manuscript, Johnson Graduate School of Management, Cornell
University (April).
Hördahl, Peter. 2008. “The Inflation Risk Premium in the Term Structure of Interest Rates.”
BIS Quarterly Review (September):23–38.
Ibbotson, Roger G., and Peng Chen. 2003. “Long-Run Stock Returns: Participating in the Real
Economy.” Financial Analysts Journal, vol. 59, no. 1 (January/February):88–98.
Ibbotson SBBI. 2011. 2011 Classic Yearbook: Market Results for Stocks, Bonds, Bills, and Inflation,
1926–2010 (Chicago: Morningstar).
Kaplan, Paul D. 2011. Frontiers of Modern Asset Allocation. Hoboken, NJ: John Wiley & Sons.
Kocherlakota, Narayana R. 1996. “The Equity Premium: It’s Still a Puzzle.” Journal of Economic
Literature, vol. 34, no. 1 (March):42–71.
Mehra, Rajnish. 2003. “The Equity Premium: Why Is It a Puzzle?” Financial Analysts Journal,
vol. 59, no. 1 (January/February):54–69.
Mehra, Rajnish, and Edward C. Prescott. 1985. “The Equity Premium: A Puzzle.” Journal of
Monetary Economics, vol. 15, no. 2 (March):145–161.
Shiller, Robert J. 2000. Irrational Exuberance. 2nd ed. Princeton, NJ: Princeton University Press.
Siegel, Laurence B. 2010. “Black Swan or Black Turkey? The State of Economic Knowledge
and the Crash of 2007–2009.” Financial Analysts Journal, vol. 66, no. 4 (July/August):6–10.
Standard & Poor’s. 2008. “S&P 500 Stock Buybacks Retreat in Q1 but Remain Strong.” Press
release (18 June).